Vertical Semiconductor, an innovative spinout from MIT, has successfully raised $11 million in seed funding to commercialize its groundbreaking vertical Gallium Nitride (GaN) transistors. This significant investment, led by Playground Global, targets the rapidly escalating power bottlenecks in AI data centers, promising remarkable improvements in energy efficiency by up to 30% and a drastic reduction in power system footprint by half.
The relentless demand for artificial intelligence capabilities is pushing data center infrastructure to its limits, with power delivery emerging as one of the most critical challenges. Conventional silicon-based power conversion solutions are struggling to keep up, generating excessive heat and consuming vast amounts of energy. This is where Vertical Semiconductor steps in, aiming to redefine how electricity is delivered to the silicon brains of AI.
The GaN Advantage: Why Vertical is a Game Changer
At the heart of Vertical Semiconductor’s solution lies Gallium Nitride (GaN), a material renowned for its superior performance over traditional silicon in power applications. GaN transistors can operate at higher voltages and frequencies, reducing energy loss and heat generation. What sets Vertical Semiconductor apart, however, is its novel vertical architecture.
Unlike most existing GaN chips where transistors are laid out horizontally, Vertical Semiconductor stacks the components of the transistor vertically. This compact design allows for:
- Higher Density and Voltage: More transistors can be packed into a smaller area, enabling higher power delivery in a reduced footprint.
- Improved Thermal Management: The vertical structure is inherently better at dissipating heat, a critical factor in maintaining performance and reliability in demanding data center environments.
- Enhanced Efficiency: By bringing energy conversion closer to the chip, power loss is significantly cut, leading to efficiency gains of up to 30%.
- Reduced Footprint: The smaller, cooler chips translate to data center power systems that are half the size, freeing up valuable rack space for more compute power.
Cynthia Liao, CEO and co-founder of Vertical Semiconductor, emphasized the broader impact, stating, “We’re not just improving efficiency, we’re enabling the next wave of innovation by rewriting how electricity is delivered in data centers at scale.”
From MIT Labs to AI Infrastructure
Vertical Semiconductor’s technology is rooted in extensive research conducted at MIT’s Palacios Group, a globally recognized leader in GaN research. Professor Tomas Palacios, a co-founder of the company, led the foundational work, which was further developed by Joshua Perozek, whose doctoral research specifically focused on this vertical GaN approach. This deep academic heritage provides a strong scientific backbone to the company’s commercial ambitions, as reported by Reuters.
The company has already demonstrated its vertical GaN technology on 8-inch wafers using standard CMOS semiconductor manufacturing methods. This crucial compatibility with existing processes means seamless integration into semiconductor manufacturing lines, paving the way for scalable and cost-effective production for devices ranging from 100 volts to 1.2 kV.
Addressing AI’s Growing Power Appetite
The surge in AI workloads places immense strain on current data center power capacity and energy budgets. As AI models grow larger and more complex, the energy required to power them increases exponentially. This creates a vicious cycle of escalating operational costs and environmental concerns due to power loss in the form of heat.
Vertical Semiconductor’s focus on efficient power delivery directly tackles this bottleneck. By making power systems more efficient and power dense, their vertical GaN transistors unlock new performance potential in AI data centers, ultimately allowing for faster, more powerful AI computations without proportionally increasing energy consumption. This is a vital step towards more sustainable and scalable AI development.
The Road Ahead: Prototypes and Commercialization
With the fresh infusion of $11 million in seed funding, Vertical Semiconductor is rapidly advancing its product roadmap. The company plans to begin early sampling of its first prototype packaged devices for customers by the end of the current year. Looking further ahead, a fully integrated solution is targeted for release in 2026.
While competitors like Transphorm, Navitas, and EPC are also making strides in GaN technology for data centers, Vertical’s unique vertical architecture and CMOS compatibility could provide a significant scaling advantage. As Matt Hershenson, a partner at lead investor Playground Global, noted, “They’re not just advancing the science – they’re changing the economics of compute.” This shift could lead to substantial long-term cost savings for data center owners and operators, making advanced AI more accessible and economically viable.
The core research originating from the cutting-edge facilities of MIT’s Palacios Group positions Vertical Semiconductor to be a pivotal player in the evolving landscape of AI infrastructure, offering a compelling next-generation solution that promises a step-wise transformation in power delivery for artificial intelligence.
